Crypto 2026: the 21 tokens that rose with the market at −40%

Of the 202 cryptocurrencies that met reasonable investment criteria on January 1, 2026, only 21 closed the semester in the green. One in ten. The group median lost 41.3%, Bitcoin fell 33.1%, and Ethereum 47.1%. The interesting question is not how much was lost—which is what everyone is talking about—but how it is possible that twenty-one tokens appreciated in such a landscape. The answer holds little mystery and quite a bit of accounting: ten of those twenty-one collect fees for protocol use or destroy supply with them. This article publishes the two complete lists—winners and losers—based on the same universe of cryptocurrencies, groups them by narratives, explains what each project is, and details why the rankings published in July 2026 show figures that look nothing like these.

How much did the market actually fall in the first half of 2026?

It depends on what is measured and, above all, how it is averaged. Across our universe of 202 tokens, the arithmetic mean for the semester was −30.0% and the median was −41.3%. There is an eleven-point difference between the two figures, and neither is miscalculated: the mean is pulled up by the tail of winners, led by a token that rose 469%. The median tells us what happened to the token right in the middle, which represents the typical investor experience. When a headline announces the average performance of a crypto sector, it is almost always describing two or three tokens and presenting it as the state of the market.

The data in the last row deserves a separate reading. Beating Bitcoin and making money were two different things: 76 tokens performed better than BTC, yet only 21 ended above zero. Fifty-five tokens lost less than the benchmark asset and still lost money. In a bearish semester, "better than the market" is still a loss.

Which tokens rose in the first half of 2026?

The complete list, ordered by performance. The January market cap determines entry into the universe; the June cap allows us to see if the rise was accompanied by real size or just price action.

This table should not be read as a list of recommendations. It is the opposite: twenty-one names out of two hundred and two, and several of them have continued to fall since June 30. dYdX closed the semester with a 152 million market cap and is worth 96 today; KAITO closed at 136 and is now at 218. Subsequent drift goes both ways and is not captured by any of the figures in the table.

What do the cryptocurrencies that rose in 2026 have in common?

Grouped by the reason for the rise rather than sector labels, the twenty-one winners fall into four very unequal groups.

Ten collect fees or destroy supply. Hyperliquid, JUST, Jito, Aerodrome, Jupiter, Injective, dYdX, and TRON generate revenue from protocol use; Rain allocates 2.5% of trading volume to buy back and burn its token; Audiera applies weekly buybacks with platform revenue. This is the largest group and the only one with an explanation that does not depend on market sentiment: while the price of everything else depends on new money entering, a protocol that generates revenue has a different floor.

One is dollar infrastructure. Stable, a chain that uses USDT as native gas. In a semester where stablecoin market capitalization continued to grow while prices fell, the rails on which that money moves appreciated even if the money they carry doesn't move from a dollar. BUILDon, which also lives in the USD1 environment, is a memecoin and counts as such.

Five live off the AI narrative. Grass, Kite, Render, KAITO, and NEAR. Here, the revenue is much more debatable than in the first group, and performance is notably lower: four of the five stayed below +21%, far from the triple digits at the top. The narrative sustained the price; it did not ignite it.

Five are pure speculation. BinanceLife, KOGE, and BUILDon in the BNB ecosystem, Terra Luna Classic—the chain that collapsed in 2022 and continues to trade—and Axie Infinity. The top-ranked asset belongs to this group, with +469%, and is the reason the market mean deviates eleven points from the median.

The distribution dismantles both common stances at once: twelve of the twenty-one winners have an identifiable economic mechanism, yet first place went to a pure speculation token, with nearly triple the rise of the runner-up.

Did the market reward tokens that return value to the holder?

This is the question that separates this semester from previous ones. For years, the token served to finance the project: it was issued, sold, and the money paid for development, with dilution that the buyer accepted in exchange for the promise. In the last two years, the opposite movement has spread—the protocol generates revenue and returns part of that cash to the token through buybacks, burns, or fee distribution—and the first half of 2026 is the first long bearish window in which this model can be measured against the other.

A criterion must be declared here before showing the result, because everything depends on it: we separate buybacks and burns financed by the protocol's own revenue from corporate burns decided by a company with its treasury, as exchange tokens have done for years. Nominally, both destroy supply. The outcome of the semester looks nothing alike.

Ten of the twenty-one winners of the semester come from a group of twenty-five cryptocurrencies out of two hundred and two. Within that group, four out of ten closed positive; outside it, seven out of a hundred. And the eleven exchange tokens, which have been burning supply quarterly for years and which any superficial reading would have lumped into the same category, ended eleven out of eleven in the red. The difference is not in destroying supply, but in who pays for the destruction: protocol usage or a company's balance sheet. While this is not a formal statistical contrast, the breadth separation between blocks is too large to be read as noise.

What the data does not say is equally important. The median of the group with its own revenue still lost 30.1%: having a buyback saved no one on its own. Included are Pendle at −30.1%, pump.fun at −26.3% despite having the most aggressive program in the sector, Aave at −41.7%, Curve at −48.7%, Uniswap at −50.8%, Lido at −58.3%, and Ethena at −64.0%. All distribute or burn, and all sank.

The precise reading, then, is that the market rewarded buybacks that weigh significantly against the token's size and was indifferent to symbolic ones. Hyperliquid buys back with fees that grew during the semester; Uniswap's fee switch—the toggle that activates fee distribution—moves a small figure relative to its market cap. This is the distinction we already measured when comparing which protocols actually return value and which just announce it, and these six months put it to the test with the market against it. A methodological warning: we composed the list of twenty-five based on public and documented mechanisms, and it is not exhaustive; it should be read as a robust indication rather than a closed census.

Which narratives worked and which ones sank?

Assigning each token a single primary narrative—our own criteria, as aggregators tag the same token with five categories at once and averaging by category would count it five times—the semester breaks down as follows. The column that matters is the last one: the mean and median tell us how much was lost, but the proportion in the green tells us if the narrative produced winners or just survivors.

DeFi was the only narrative that produced winners regularly: nine out of forty-seven, nearly one in five, and the best median in the table after filtering for noise. This coincides with the group where fee distribution lives, which is the same story told from another angle.

Infrastructure was the disaster of the semester, and it is the largest block: fifty-four tokens, a median of −46.3%, and only two in the positive, one of them for a very specific reason unrelated to being a chain. The thesis of buying the emerging L1, which dominated the previous two years, wiped out Sonic, Aptos, Flow, Walrus, and Bitcoin Cash all at once.

Two narratives ended with zero winners. Exchange tokens, eleven out of eleven in the red despite their burn programs. And privacy, ten out of ten, with Monero losing 30% and Zcash 22% in a year of regulatory pressure where the thesis seemed to have every argument in its favor. That an entire narrative fails to produce a single winner in six months is a more powerful data point than any median.

AI landed in an intermediate spot, less brilliant than the hype suggests: 33 of the 202 tokens, one-sixth of the sample, fifteen percent in the green, and a median of −38.1%—meaning better than the market but far from being a safe haven. None of its winners reached +60%.

Did being a large-cap token protect against the 2026 fall?

This is the most widespread belief about bear markets, and it did not hold true this semester. Dividing the universe into tiers by January 1 market capitalization, the medians are almost identical among the top hundred.

Between the top 20 and the 51-100 tier, there is only a three-point difference in the median. Only the long tail, from rank 101 downward, fell appreciably worse, and even then the gap is eight points, not the abyss usually described. The probability of picking a winner also did not improve with size: between 7% and 12% across all tiers. Bitcoin Cash, which started the year with nearly 12 billion in market cap, lost 66.7%.

What were the worst cryptocurrencies of 2026 and why don't they appear in other rankings?

Almost no ranking publishes this table. The CoinLore page dedicated to yearly performance, which is the closest reference to what we do here, only shows coins with positive performance: losers are counted in the breadth statistics but do not appear in the list. The result is a market snapshot where only the winners are visible.

There is a methodological detail that makes this list more benevolent than it should be. Our universe is built on today's top 500 coins by market cap, whose floor is at 42 million dollars. A token that was worth 150 million in January and has since fallen below that floor does not appear here because it is no longer in the starting list. This bias only goes one way: the losers' table is incomplete at the bottom, and the actual losses for the semester were worse than what we publish. The floor of the starting list is 42 million dollars: anything that has fallen below that since January is excluded from this table.

Why do other rankings give such different figures?

Due to three methodological decisions that are rarely explained and that change the result completely.

The first is averaging with the mean. CoinLore's annual performance page goes as far as publishing an equal-weighted average return of +824.43% for the top 100, consulted on August 4, 2026. This is not a calculation error: it is what happens when you equally average coins that multiplied by eight hundred with coins that lost half their value. Our own universe, with a median of −41.3%, has a mean of −30.0% due to the same mechanism on a smaller scale. Any ranking that gives a single average performance figure without the median alongside it is describing its tail, not its market. It is worth noting that the same page, in its breadth statistic, states that only 10% of the coins it tracks closed positive —55 gainers out of 554—: a figure that almost exactly matches our 10.4%, despite starting from a universe built on different criteria.

The second is the entry perimeter. If the filter is applied to today's market cap, everything that crashed is automatically excluded: projects worth a billion in January that are worth fifty million today no longer make the cut and disappear from the sample. The list becomes full of survivors. We apply the cut using January 1 data, which is the only way to measure what happened to a real portfolio from that time.

The third is market capitalization itself. Rain illustrates the problem better than any explanation: the same token, on the same day, appears at rank 14 on CoinGecko with about 8.89 billion in market cap and at rank 201 on CoinMarketCap. And the detail that makes it baffling: both show practically the same market cap —$8.653 billion on CoinGecko, $8.6519 billion on CoinMarketCap— and the same circulating supply, 694,586 million tokens. CoinMarketCap publishes no note explaining why it does not factor that figure into its ranking, so the cause is unknown to us; the fact is verifiable on both pages. A two-hundred-rank gap for the same asset, on the same day, with no price discrepancy. Anyone building a ranking by market cap without stating which provider gives the circulating supply and by what criteria they validate it is publishing that provider's opinion.

There is a fourth warning, this one of a practical nature: there are repeated tickers in our universe. DAI, SAFE, USDF, and HOLO designate more than one distinct token within the same top five hundred. Reading a ranking by symbol rather than by project is an easy way to buy the wrong thing.

What is each of the 21 cryptocurrencies that rose?

Brief profiles of the twenty-one winners, with their official websites so anyone can verify for themselves. Outbound links in this section carry the nofollow attribute: appearing in a performance table is not an editorial endorsement by CleanSky of any project.

BinanceLife (币安人生), +469.3%. Speculation token from the BNB Chain ecosystem, listed on Binance on January 7, 2026, with no official website declared in aggregators. It is the semester's biggest winner and also the example of why the market mean deviates from the median.

Stable (STABLE), +178.7%. Layer 1 chain backed by Bitfinex and Tether that uses USDT as native gas, allowing users to operate without holding a separate volatile asset. Its token is for network governance and staking. stable.xyz

Hyperliquid (HYPE), +155.4%. Perpetuals order book—futures without an expiration date—with its own chain, and the largest by volume in the sector. It collects fees on every trade and allocates part to buy back its token, a circuit we analyzed in detail in the piece on its builder codes. hyperliquid.xyz

JUST (JST), +108.7%. A suite of DeFi protocols on TRON—lending, collateralized stablecoin, and swaps—whose token captures fees from the chain ecosystem that generates the most revenue from usage. just.network

Audiera (BEAT), +98.1%. Rhythm and AI music creation game on BNB Chain, featuring two virtual vocalists and a revenue-sharing model. It applies periodic token buybacks using platform earnings. audiera.fi

Rain (RAIN), +96.3%. Prediction market protocol on Arbitrum that allocates 2.5% of trading volume to buy back and burn its token. It is also the case that best illustrates the disagreement between aggregators: CoinGecko ranks it 14th and CoinMarketCap 201st. rain.one

Jito (JTO), +81.0%. Liquid staking and MEV capture infrastructure—the value extracted by reordering transactions—on Solana. Its governance approved allocating the entire DAO fee portion to token buyback and burn. jito.network

Grass (GRASS), +59.1%. DePIN network—physical infrastructure coordinated with on-chain incentives—on Solana that pays users for sharing their unused residential bandwidth, which AI labs use to crawl the public web at scale. It claims about two and a half million nodes spread across 190 countries. grass.io

BUILDon (B), +58.4%. Memecoin from the BNB Chain ecosystem, presented as the chain's mascot, whose narrative revolves around providing liquidity to the USD1 stablecoin and the World Liberty Financial environment. buildon.online

KOGE (KOGE), +45.5%. Token of the 48 club on BNB Chain, part of the block of projects promoted by the Binance Alpha program. Speculative by nature with no declared revenue mechanism. 48.club

Terra Luna Classic (LUNC), +41.8%. The original Terra chain, the one that collapsed in May 2022 dragging down UST. It continues to trade and maintains a community that votes on supply burns; its appreciation is speculative and does not respond to new economic activity. terra-classic.io

Kite (KITE), +20.7%. Payments infrastructure for artificial intelligence agents, a field we covered when analyzing agentic payments and the x402 standard. gokite.ai

Axie Infinity (AXS), +19.2%. The game that defined the play-to-earn model in 2021, whose token is for governance and staking. Its rise in a semester where the rest of its narrative fell by a median of 47.8% makes it the exception in its block. axieinfinity.com

Aerodrome Finance (AERO), +18.5%. The largest decentralized market on the Base network, with a vote-escrowed model that distributes fees among those who lock the token. aerodrome.finance

NEAR Protocol (NEAR), +17.7%. Layer 1 chain that reoriented its discourse and much of its development toward artificial intelligence. Although it is a Layer 1 chain, in this analysis it counts toward the AI narrative: it is that thesis, and not the infrastructure one, that sustained its price. near.org

Render (RENDER), +17.1%. Distributed graphics computing network that connects those with idle graphics cards with those who need to render or train models. renderfoundation.com

Jupiter (JUP), +12.3%. Solana swap aggregator: it routes every trade through the market offering the best price and charges for doing so. It concentrates a very high portion of the chain's swap volume. jup.ag

KAITO (KAITO), +10.8%. Platform that applies AI to organize information and attention in the crypto market, with analysis products and attention markets. kaito.ai

TRON (TRX), +10.8%. Layer 1 chain with the highest stablecoin transfer volume in the sector, translating into sustained fee revenue independent of the speculative cycle. tron.network

Injective (INJ), +8.1%. Chain oriented toward financial applications that holds a weekly auction where tokens collected from application fees are burned. injective.com

dYdX (DYDX), +6.6%. One of the veteran perpetuals order books in the sector, with its own chain and fee distribution to stakers. It closed the semester in the green and has fallen sharply since then. dydx.trade

How was this ranking constructed?

The period is closed: from January 1 to June 30, 2026, with opening and closing prices taken from the historical series of the CoinGecko public API. Today's market cap figures appearing in the text correspond to August 4, 2026, and are noted as such; they are not mixed with the semester's returns.

The starting universe consists of the top 500 coins by market cap, to which three filters are applied:

1. Minimum market capitalization of 100 million dollars on January 1, 2026, not today. This decision avoids survivorship bias at the top: it measures what happened to what was investable at the start of the year.
2. Median daily volume of at least one million dollars during the semester. Market cap without volume is a price at which no one can exit. Of ranks 251 to 500 on today's list, more than half do not reach this threshold.
3. Exclusion of stablecoins, wrapped assets, and liquid staking derivatives. These are not independent bets: they replicate the value of something else. Including a flat stablecoin in a winners' ranking, in a semester where the market falls 40%, would make it the year's big winner for doing nothing. The same applies to tokenized yield instruments—treasury bonds, mortgage credits—which appeared in the raw sample with rises of 1.5% and 2.4% that are, simply, their coupon.

Of the initial 500, 202 pass the three filters. The median, mean, and breadth figures in this article are always calculated on those 202, and the winners' and losers' tables come from the same set, using the same criteria and without discretionary exclusions.

The two known limitations are stated: the universe starts from today's top 500, so anything that has fallen below rank 500 since January is not captured and the losers' list falls short; and the assignment of each token to a narrative is our editorial decision, not an official classification, because aggregator category systems assign multiple labels to the same token and averaging by category would count it more than once.

Sources and links: CoinGecko — Public API for historical prices and market cap (accessed Aug-4-2026) · CoinLore — Performance by calendar year (positive coins only) · CoinGecko — Winners and losers by rolling window · CoinMarketCap — Winners and losers · CoinMarketCap — Rain profile (rank 201) · CoinGecko — Rain profile (rank 14)